Add CCPMF dataset

DeepSpeech/Dockerfile.train

@@ -90,6 +90,7 @@ RUN apt-get -qq update && apt-get -qq install -y --no-install-recommends \
     curl \
     wget \
     git \
+    ffmpeg \
     python3 \
     python3-pip \
     ca-certificates \
@@ -175,6 +176,9 @@ RUN pip install parso==0.7.0
 
 RUN python setup.py install
 
+# For CC PMF importer
+RUN pip install num2words
+
 WORKDIR $HOMEDIR
 
 ENV PATH="$HOMEDIR/kenlm/build/bin/:$PATH"

DeepSpeech/checks.sh

@@ -27,6 +27,8 @@ done;
 
 mkdir /mnt/extracted/data/ || true
 
+python -c "import tensorflow as tf; tf.test.is_gpu_available()"
+
 # Checking with basic LDC93S1 before running into heavy-load
 pushd $HOME/ds/
 	./bin/run-tc-ldc93s1_new.sh 2 16000

DeepSpeech/fr/import_ccpmf.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+
+set -xe
+
+pushd $HOME/ds/
+	if [ ! -f "/mnt/extracted/data/ccpmf/ccpmf_train.csv" ]; then
+		#  Hot patching like that.
+		sed -ri 's/MAX_SECS = .*/MAX_SECS = 4.5/g' bin/import_ccpmf.py
+
+		python bin/import_ccpmf.py \
+			${IMPORTERS_VALIDATE_LOCALE} \
+			/mnt/extracted/data/ccpmf/
+	fi;
+popd

DeepSpeech/fr/importers.sh

@@ -11,3 +11,5 @@ import_trainingspeech.sh
 import_slr57.sh
 
 ../import_m-ailabs.sh
+
+import_ccpmf.sh

DeepSpeech/fr/metadata.sh

@@ -4,7 +4,7 @@ set -xe
 
 export METADATA_AUTHOR="DeepSpeech-FR-Team"
 export METADATA_MODEL_NAME="deepspeech-fr"
-export METADATA_MODEL_VERSION="0.5"
+export METADATA_MODEL_VERSION="0.6"
 export METADATA_CONTACT_INFO="https://discourse.mozilla.org/c/voice/fr"
 export METADATA_LICENSE="MIT-0"
 export METADATA_LANGUAGE="fr-FR"